作者位次 #7/8ACL 2026Poster· Natural Language Processing
Detecting RAG Extraction Attack via Dual-Path Runtime Integrity Game
Yuanbo Xie · Yingjie Zhang · Yulin Li · Shouyou Song · Xiaokun Chen · Zhihan Liu · Liya Su · Tingwen Liu
摘要
Retrieval-Augmented Generation (RAG) systems augment large language models with external knowledge, yet introduce a critical security vulnerability: RAG Knowledge Base Leakage, wherein adversarial prompts can induce the model to divulge retrieved proprietary content. Recent studies reveal that su...